pnoise spurs at multiples of clock frequency.

Vamshi,

Can you post your Netlist (the input.scs) here and also give the spectre subversion you're using? (Type "spectre -W" at the unix prompt; it will also appear near the top of the spectre output log)?

Whilst I can throw together a simple example to show this, seeing your setup will help to understand why you are getting this.

Thanks,

Andrew 

I tried your circuit (with two pnoise analyses in the netlist to make it simpler), and with some made up models and device sizes, and I don't see the problem. Spectre also reports (using the same version you're using):

 Warning from spectre at freq = 100 MHz during PNoise analysis `pnoise'.
    WARNING (SPCRTRF-15037): Infinite flicker noise is ignored. 

Here's what I used (I fixed your line wrapped netlist - I assume you're using Chrome, which causes some problems with the forums on this web site, as far as I know).

 // Generated for: spectre
// Generated on: Dec 19 12:31:13 2013
// Design library name: xxxx
// Design cell name: test_pss_pnoise
// Design view name: schematic
simulator lang=spectre
global 0
//include "xxxxx.scs"
model nch_lvt bsim4 type=n
model pch_lvt bsim4 type=p
// Library name: xxxxxxxxx
// Cell name: CKINVX2
// View name: schematic
subckt CKINVX2 VDD VSS i o
MN0 (o i VSS VSS) nch_lvt w=1u l=0.1u
MP0 (o i VDD VDD) pch_lvt w=2u l=0.1u
ends CKINVX2

// End of subcircuit definition.
// Library name: xxxx
// Cell name: test_pss_pnoise
// View name: schematic
V1 (avss 0) vsource dc=0 type=dc
V0 (avdd0p8 avss) vsource dc=800.0m type=dc
I1 (avdd0p8 avss in out) CKINVX2
V2 (in avss) vsource type=pulse val0=0 val1=800.0m period=20n delay=10p \
  rise=10p fall=10p width=10n
C0 (out avss) capacitor c=10f

simulatorOptions options reltol=1e-3 vabstol=1e-6 iabstol=1e-12 temp=27 \
 tnom=27 scalem=1.0 gmin=1e-12 rforce=1 maxnotes=5 maxwarns=5 digits=5 \
 cols=80 pivrel=1e-3 sensfile="../psf/sens.output" checklimitdest=psf
pss pss fund=50M harms=10 errpreset=conservative
+ annotate=status
pnoise ( out avss ) pnoise start=1k stop=500M dec=20
+ maxsideband=10 annotate=status
pnoiseAlt ( out avss ) pnoise start=1k stop=501M step=1.7M
+ maxsideband=10 annotate=status
modelParameter info what=models where=rawfile
element info what=inst where=rawfile
outputParameter info what=output where=rawfile
designParamVals info what=parameters where=rawfile
primitives info what=primitives where=rawfile
subckts info what=subckts where=rawfile
save out avss
saveOptions options save=selected

Regards,

Andrew.

 Thanks for your efforts Andrew, I now need to understand if the issue is with models used or some kind of env variable which differs from your test.

please suggest.

-Vamshi

It seems unlikely that it would be an environment variable. I can only imagine that it might be model related. Maybe the flicker noise profile is such that the flicker noise is still high near the harmonic multiples.

That said, I just tried setting dec=200 on the pnoise line, and I no longer get the message about infinite flicker noise, and there is a huge peak at 100MHz if I do this. So my guess is that due to floating point numerical error, it may not detect the near-infinite flicker noise and out a value instead (e.g. if it was a small fraction of a Hertz away from the harmonic multiple).

Either way, going through exact (or very near exact) multiples of the fundamental probably doesn't make a lot of sense anyway.

Regards,

Andrew.

OK, then I am only left with carefully choosing the step sizes/ sweep points. Linear step size with an irrational number does the job but actually slows my sims on actual circuit chargepump for PLL. (ofcourse still see a bit of wierd behaviour at high frequencies)

Do you have any better suggestions to avoid those near/exact multiple frequencies in sims.

Thanks,

Vamshi

Vamshi,

Could you alter your input frequency to not be exact "decade" frequency? Then you could use a number of points per decade, and know that your harmonics don't coincide with 10M, 100M, 1G etc.

Andrew.

 I understand that, however input frequncy is a system level spec and cant be changed.

Thanks,

Vamshi

The other alternative is that you specify start, stop and log - this is the number of log steps beween the two end points. For example:

pnoiseLog ( out avss ) pnoise start=1k stop=501M log=110
+ maxsideband=10 annotate=status

So I moved the end point to not be on an exact multiple, and I made the log roughly 20*numDecades (assuming you had dec=20 before).

Regards,

Andrew.